PAGES
← all tools
◆ SAST / SCA / Code Security

VulnHunter

v0.1.0 open-source

VulnHunter is an open-source agentic AI security tool that applies attacker-first analysis to source code to identify exploitable vulnerabilities.

source ↗ rss ↗ What is a cyber toolchain? last repo push Jul 18, 2026

Summary

VulnHunter is an open-source, agentic AI security tool that runs as a set of Claude Code skills for finding and fixing exploitable vulnerabilities in source code. Rather than flagging suspicious patterns like a passive SASTSASTStatic application security testing, analyzing source code or compiled code for vulnerabilities without running the application. It helps developers find problems before deployment. scanner, it works forward from attacker-accessible entry points, tracing paths to dangerous sinks and filtering results through a multi-stage falsification pipeline before proposing fixes, so it targets application security and appsec engineering teams who want fewer false positives than conventional static analysis. It requires Opus-class model access supplied by the user, and can run headless in CI/CD pipelines via its agent runtime, or in batch mode across many repositories using its harness tooling. Developed internally at Capital One before release to the community, it is a young project — first committed in 2026, with four contributors and one release so far, though that release landed just over a month ago.

VulnHunter is an open-source agentic AI security tool that applies attacker-first analysis to source code to identify exploitable vulnerabilities.

How VulnHunter helps

Which entry points in this codebase could an attacker actually reach, and where do they lead?

starts from APIs, network messages, and file uploads and traces each one forward through the code to the dangerous sinks it can reach, rather than starting from the sink and guessing backward

Which of these findings are real, exploitable bugs rather than noise?

pushes every candidate through a multi-stage pipeline — recon, parallel hunting, adversarial disproof, and a capability filter — before it's reported, so what survives has been actively argued against, not just pattern-matched

Is this vulnerability actually fixed, or does the patch just look right?

a separate read-only agent re-runs the original exploit path against the patched code and issues its own pass/fail verdict, independent of whichever process produced the fix

How do I get from a confirmed finding to a mergeable fix without doing it by hand?

writes a failing security test that demonstrates the exploit, implements the smallest fix that turns it green, confirms the exploit is now blocked, and opens a pull request with that evidence attached

Which of the repositories I'm responsible for actually have exploitable issues, not just which ones have code smells?

runs the same attacker-first scan across a whole list of repositories, resuming where a previous batch left off, and centralizes every confirmed finding for review in one place

How accurate is this against vulnerabilities I already know are there?

the benchmarking harness runs the scanner against a corpus of known-vulnerable code and scores detection accuracy with an LLM judge, giving a measured baseline instead of a claimed one

all 5 features, with the evidence for each →

Languages

Not documented

Features

5 capabilities

Built from VulnHunter's releases, docs, README and our own runs. “Seen” marks the month we first spotted a release the vendor didn't date. Group names and tags are our labels.

Compare with peers →

  • VulnHunter covers 2 of the 8 vulnerability assessment methods capabilities we track.

What VulnHunter offers you

  • CI/CD pipeline integration1
  • Ticketing integration1
  • Command-line interface
All capabilities 5 capabilities
Attacker-path vulnerability scanning shipped Scans a codebase by mapping attacker entry points to dangerous sinks and filters findings through a multi-stage falsification pipeline to reduce false positives. 2 releases · 3 other sources · first seen Jul 2026

Shared capability: Static application security testing

release

  • Initial release of VulnHunter as an open-source agentic AI security tool with CLI capabilities. v0.1.0 · Jul 2026 · source · release history
  • Adds /vulnhunt Claude Code skill: entry-point-forward scanner that maps attacker paths to dangerous sinks and filters findings through a multi-stage falsification pipeline (Recon → Parallel Hunt → Adversarial Disprove → Capability Filter). v0.1.0 · Jul 2026 · source · release history

docs

  • Agentic AI cyber tool that performs attacker-first source code analysis to identify exploitable vulnerabilities, map attack paths, and generate evidence-backed

example

  • Run the full attacker-first vulnerability scan against a codebase inside a Claude Code session. claude --model opus --add-dir ~/.claude/skills/vulnhunt --add-dir ~/.claude/skills/vulnhunt/phases # Then inside the Claude Code session: /vulnhunt v0.1.0 · Jul 2026 · source
  • Launch scanner session with skills loaded claude --model opus --add-dir ~/.claude/skills/vulnhunt --add-dir ~/.claude/skills/vulnhunt/phases
Batch scanning and benchmarking shipped suggested Manages a list of repositories for batch scanning with resume support, centralises collected findings, and evaluates detection accuracy against a known-vulnerable corpus using LLM-judge scoring. 1 release · 4 other sources · first seen Jul 2026

Suggested capability: Static application security testing. This is our guess, not confirmed.

release

  • Adds harness/ batch-scanning and benchmarking tooling: manage a repo list, resume interrupted scans, collect findings centrally, and evaluate detection accuracy against a known-vulnerable corpus with LLM-judge scoring. v0.1.0 · Jul 2026 · source · release history

example

  • Batch-scan a list of repositories and collect all findings for centralized review, skipping repos already processed. cd harness python -m local_harness.batch.run scan --resume python -m local_harness.batch.run collect v0.1.0 · Jul 2026 · source
  • Clone and scan every repo in list python -m local_harness.batch.run scan
  • Resume batch scan, skipping processed repos python -m local_harness.batch.run scan --resume
  • Execute full benchmark run python -m local_harness.benchmark.run
Fix verification shipped Runs a read-only agent that independently validates whether a prior remediation successfully blocked the exploit and emits a per-finding verdict. 1 release · 2 other sources · first seen Jul 2026

release

  • Adds /vulnhunt-fix-verify skill: independent, read-only agent that validates whether a finding was successfully remediated and emits a per-finding verdict. v0.1.0 · Jul 2026 · source · release history

example

  • Independently verify that a prior fix actually blocked the exploit before merging, using the read-only verifier agent. claude --model opus --add-dir ~/.claude/skills/vulnhunt-fix-verify \ --add-dir ~/.claude/skills/vulnhunt-fix-verify/phases # Then inside the Claude Code session: /vulnhunt-fix-verify repo=<abs_path>… v0.1.0 · Jul 2026 · source
  • Launch fix verifier session claude --model opus --add-dir ~/.claude/skills/vulnhunt-fix-verify \
Test-driven vulnerability remediation shipped Remediates confirmed findings by writing an exploit demo, creating a failing security test, implementing a fix, verifying the exploit is blocked, and opening a reviewable PR. 1 release · 1 other source · first seen Jul 2026

release

  • Adds /vulnhunter-fix skill: test-driven remediation loop that writes an exploit demo, creates a failing security test (RED), implements the fix (GREEN), verifies the exploit is blocked, and opens a reviewable PR. v0.1.0 · Jul 2026 · source · release history

example

  • Launch fixer session with skills loaded claude --model opus --add-dir ~/.claude/skills/vulnhunter-fix
CI/CD headless scanning shipped Runs the vulnerability scanner autonomously in a headless runtime suitable for CI/CD pipelines, publishing results and opening GitHub issues for confirmed findings. 1 release · first seen Jul 2026

Shared capability: Web vulnerability scanning

Also counts toward: CI/CD pipeline integration, Ticketing integration

release

  • Adds vulnhunter-agent/ headless runtime: clones targets, runs /vulnhunt, publishes results, and opens GitHub issues for confirmed vulnerabilities — suitable for CI/CD pipelines. v0.1.0 · Jul 2026 · source · release history

Lines in monospace are the tool's own words: help text parsed from its source, or an endpoint from its API document. Under release, a line is our summary of a dated release. Under docs, it is quoted from the tool's own documentation or website. Each links back to the source it came from.

Score card

  1. Shipping velocity releases a month, last 12 months
  2. Documentation quality release-notes rubric, 0–100
  3. Breaking-change discipline how rarely a release breaks users, and notice given
  4. Maintenance risk quiet time against its own cadence, maintainers
  5. Innovation & surface new capabilities, API and MCP surface
  6. Community health
  7. Contributor trend distinct committers, this year against last
  8. Maintainer concentration bus factor now against a year ago
  9. Issue backlog trend issues closed per issue opened, 12 months
  10. Responsiveness median time to a maintainer’s first reply

Nine scores for VulnHunter, each built from the releases, notes and API history this site tracks — and, on Researcher and Business, where it stands among every measured tool in its category and how that has moved.

Score cards are part of every paid plan.

See plans →

Each score is 0–100 on a published rubric, derived from data this site already shows. A dimension we cannot measure says so — it is never scored 0 and never enters a rank.

How this is scored
  • Shipping velocity — releases a month over the last 12 months, on a log scale where 16 a month scores 100. Needs 3 releases in the window.
  • Documentation quality — the release-notes rubric published on the analytics page, unchanged.
  • Breaking-change discipline — the share of releases carrying a breaking change (none scores 100, half or more scores 0). Only projects that label breaking changes are measured: a clean record cannot be told from an unlabelled one.
  • Maintenance risk — quiet time measured against the tool's own release cadence (quiet longer than 90 days and three typical gaps scores 0), plus the number of active maintainers and whether contributors are growing, where known. Higher is lower risk.
  • Innovation & surface — new capabilities in the feature ledger this year, API and MCP operations added, and the share of commits that are features: the mean of whichever we hold.
  • Community health (open-source tools only) — each is a direction, not a size: contributors this year against last, the bus factor now against a year ago, and issues closed per issue opened, where steady scores 50, doubled 100 and halved 0; plus the median time to a maintainer's first reply (an hour scores about 100, a month or more 0). A commercial card has no repository and carries none of the four.
  • Documented delivery (commercial tools) — when a vendor's documented new capabilities have slowed against its own baseline, that counts against the overall standing as heavily as two low dimensions. Steady delivery, or too little history to judge, leaves the standing alone; it is never a sixth score.
  • Open-source tools are scored on repository signals; commercial tools on their published changelog. Ranks compare tools within one category and one profile and need at least five measured peers; the overall standing needs three ranked dimensions. Best-in-breed or consolidator is a shape, not a quality, and is never scored.

Release-note detail over time

See how much detail VulnHunter puts in its release notes, quarter by quarter, and compare it with tracked tools in the same category. This history uses up to ten retained, reviewed notes per tool; sparse or blank older quarters may reflect that limit.

Checking release-note history…

Explore the full chart →

Releases

2 entries

  1. July 20262

    2 entries

    1. v0.1.0 Jul 18, 2026 · issue 019

      Capital One open-sources VulnHunter: agentic AI SASTSASTStatic application security testing, analyzing source code or compiled code for vulnerabilities without running the application. It helps developers find problems before deployment. with attacker-first analysis, falsification engine, and automated fix-verify loop.

      └──▷ GET THIS VERSION 
      $ git clone --branch v0.1.0 https://github.com/capitalone/VulnHunter.git
# already have the repo? check out this version:
$ git checkout v0.1.0
      └──▷ TRY IT
      Run the full attacker-first vulnerability scan against a codebase inside a Claude Code session.
      $ claude --model opus --add-dir ~/.claude/skills/vulnhunt --add-dir ~/.claude/skills/vulnhunt/phases
# Then inside the Claude Code session:
/vulnhunt
      Independently verify that a prior fix actually blocked the exploit before merging, using the read-only verifier agent.
      $ claude --model opus --add-dir ~/.claude/skills/vulnhunt-fix-verify \
       --add-dir ~/.claude/skills/vulnhunt-fix-verify/phases
# Then inside the Claude Code session:
/vulnhunt-fix-verify repo=<abs_path> report=<abs_path> fixed=VULN-001,... out=<abs_path>
      Batch-scan a list of repositories and collect all findings for centralized review, skipping repos already processed.
      $ cd harness
python -m local_harness.batch.run scan --resume
python -m local_harness.batch.run collect
      • Adds /vulnhunt Claude Code skill: entry-point-forward scanner that maps attacker paths to dangerous sinks and filters findings through a multi-stage falsification pipeline (Recon → Parallel Hunt → Adversarial Disprove → Capability Filter).
      • Adds /vulnhunter-fix skill: test-driven remediation loop that writes an exploit demo, creates a failing security test (RED), implements the fix (GREEN), verifies the exploit is blocked, and opens a reviewable PR.
      • Adds /vulnhunt-fix-verify skill: independent, read-only agent that validates whether a finding was successfully remediated and emits a per-finding verdict.
      • Adds vulnhunter-agent/ headless runtime: clones targets, runs /vulnhunt, publishes results, and opens GitHub issues for confirmed vulnerabilities — suitable for CI/CD pipelines.
      • Adds harness/ batch-scanning and benchmarking tooling: manage a repo list, resume interrupted scans, collect findings centrally, and evaluate detection accuracy against a known-vulnerable corpus with LLMLLMA large language model, a neural network trained on large amounts of text to recognize patterns and generate language. Applications use it for tasks such as answering, summarizing and writing code.-judge scoring.

      5 details in Everything

    2. v0.1.0 Jul 18, 2026 · issue 004

      Capital One open-sources VulnHunter: agentic AI SASTSASTStatic application security testing, analyzing source code or compiled code for vulnerabilities without running the application. It helps developers find problems before deployment. with attacker-first analysis, falsification engine, and automated fix-verify loop.

      └──▷ GET THIS VERSION 
      $ git clone --branch v0.1.0 https://github.com/capitalone/VulnHunter.git
# already have the repo? check out this version:
$ git checkout v0.1.0
      └──▷ TRY IT
      Run the full attacker-first vulnerability scan against a codebase inside a Claude Code session.
      $ claude --model opus --add-dir ~/.claude/skills/vulnhunt --add-dir ~/.claude/skills/vulnhunt/phases
# Then inside the Claude Code session:
/vulnhunt
      Independently verify that a prior fix actually blocked the exploit before merging, using the read-only verifier agent.
      $ claude --model opus --add-dir ~/.claude/skills/vulnhunt-fix-verify \
       --add-dir ~/.claude/skills/vulnhunt-fix-verify/phases
# Then inside the Claude Code session:
/vulnhunt-fix-verify repo=<abs_path> report=<abs_path> fixed=VULN-001,... out=<abs_path>
      Batch-scan a list of repositories and collect all findings for centralized review, skipping repos already processed.
      $ cd harness
python -m local_harness.batch.run scan --resume
python -m local_harness.batch.run collect
      • Adds /vulnhunt Claude Code skill: entry-point-forward scanner that maps attacker paths to dangerous sinks and filters findings through a multi-stage falsification pipeline (Recon → Parallel Hunt → Adversarial Disprove → Capability Filter).
      • Adds /vulnhunter-fix skill: test-driven remediation loop that writes an exploit demo, creates a failing security test (RED), implements the fix (GREEN), verifies the exploit is blocked, and opens a reviewable PR.
      • Adds /vulnhunt-fix-verify skill: independent, read-only agent that validates whether a finding was successfully remediated and emits a per-finding verdict.
      • Adds vulnhunter-agent/ headless runtime: clones targets, runs /vulnhunt, publishes results, and opens GitHub issues for confirmed vulnerabilities — suitable for CI/CD pipelines.
      • Adds harness/ batch-scanning and benchmarking tooling: manage a repo list, resume interrupted scans, collect findings centrally, and evaluate detection accuracy against a known-vulnerable corpus with LLMLLMA large language model, a neural network trained on large amounts of text to recognize patterns and generate language. Applications use it for tasks such as answering, summarizing and writing code.-judge scoring.

      5 details in Everything

    Project health

    Issue, contributor and maintainer activity in capitalone/VulnHunter, 2023-10 – 2026-09, as of 2026-09-25. Twelve-month figures first; the trend lines run the whole window.

    1. Contributors 7 +7 vs prior 12 months
    2. Maintainers 2 +2 joined · −0 left, last 12 months
    3. Bus factor 1 committers behind half the commits
    4. Community-filed 100% of issues opened, last 12 months
    5. Median time to close 11 days first maintainer reply 11 days
    6. Open backlog 3 of 4 issues ever filed
    1. Issues opened / month 1 · 0 communitymaintainers
    2. Open backlog 3
    3. Contributors / month 0
    4. Commits / month 0
    5. Stars 1,008
    Issue themes, last 12 months
    • other 100%

    Counts from the project's public repository and git history; no individuals are identified. Maintainers are the committers covering 80% of the last year's commits. how we measure this

    Add the badge to your README

    Maintain this project? The badge links readers straight to this page, so they can follow every release here.

    Tracked by Cyber Toolchain Choose a badge style and copy the code →
    compare.md

      Free the public facts, no sign-in Any paid plan all other sections Researcher & Business pricing history and category rank

      my-toolchain — 0 tools

        Betas, release candidates and previews. Turn it off to see final releases only. It also applies to changes you ask for over MCP when you're signed in.
        browse all tools → score card for your toolchain →
        cyber-toolchain / feedback

        Send us feedback